El Niño‐Southern Oscillation Modulation of Springtime Diurnal Rainfall Over a Tropical Indian Ocean Island

This study analyzes the interannual variability of boreal springtime (March and April) diurnal rainfall (hereafter SDR) over the second biggest Island of the Indian Ocean, Sri Lanka (SL), and expresses the possible effects of the El Niño Southern Oscillation (ENSO) events in modulating this variability. The analysis is primarily based on high spatiotemporal resolution satellite precipitation estimates and reanalysis data from 2001 to 2019. Results indicate that the SDR in SL exhibited a consistent afternoon peak throughout the study period. In contrast, the SDR's amplitude consists of a notable 2 to 4‐year oscillation period, similar to the oscillation period of ENSO events. Further analysis revealed that the interannual variation of the SDR's amplitude in SL has a negative relationship with the springtime and previous winter sea surface temperature changes in the central and eastern tropical Pacific Ocean covering NINO3.4 and NINO3 regions (referred to as ENSO indices). When there is a winter La Niña (El Niño), the following SDR in SL is more (less) active. A possible explanation is that during La Niña (El Niño) years, the region around SL experienced an enhanced (suppressed) cyclonic circulation, enhanced (suppressed) large‐scale moisture flux convergence, and enhanced (suppressed) local‐scale diurnal varying moisture flux convergence, leading to an increase (decrease) in the potential for rainfall and more (less) active SDR over SL. These findings highlight the potential to use ENSO indices to predict the interannual variation of SDR activities over SL.

spatiotemporal resolution data to study rainfall variability is a long-standing issue in the region, particularly sub-daily data over SL (Alahacoon & Amarnath, 2022;Sunilkumar et al., 2015). Therefore, earlier studies examining the Springtime Diurnal Rainfall (hereafter SDR) over SL mainly focused on the climatological feature and less on the long-term variability. The availability of satellite precipitation estimate data products with high spatiotemporal resolutions has made the sub-daily rainfall analysis possible over the data-sparse regions. Various past studies have frequently utilized satellite precipitation estimate data to study rainfall and other atmospheric variables across the globe (Huang & Chan, 2012;Kerns & Chen, 2020;Lee et al., 2021;Li et al., 2016;Murali Krishna et al., 2017;Sharma et al., 2020;Varikoden et al., 2011). Among the various products of satellite precipitation estimates, the products of Tropical Rainfall Measurement Mission (TRMM) and the Global Precipitation Measurement Mission (GPM) are the two important missions that provide sub-daily rainfall variation for studying the diurnal rainfall formation over various regions (Filho et al., 2022;Hsu et al., 2021;Huang et al., 2020;Murali Krishna et al., 2017;Sharma et al., 2020). The availability of TRMM and GPM data allows us to examine the interannual variation of diurnal rainfall in SL, which motivates this study.
The El Niño/Southern Oscillation (ENSO) is one of the climate modes that can modulate the interannual variations of rainfall over the Indo-Pacific regions through the east-west oscillation of the tropical Walker circulation (Huang et al., 2019;Krishnamurthy & Goswami, 2000;Kumar et al., 1999). During El Niño (La Niña) events, a divergence (convergence) zone at low-levels frequently appears over the western Pacific towards South Asia, resulting in drier (wetter) conditions over South Asia (Rasmusson et al., 1999;Verma & Bhatla, 2021). The El Niño (La Niña) events can also modulate the southward (northward) movement of the local Hadley circulation over the Indian Ocean, which can subsequently delay (advance) the monsoon onset process (Choudhury et al., 2021). Other studies have indicated that the interannual variation of monthly mean rainfall in SL can be modulated by the ENSO events (Jayawardene et al., 2015;Kane, 1998;Suppiah, 1996;Vialard et al., 2011;Zubair et al., 2008). For example, Suppiah (1996) and Kane (1998) examined the response of SL's rainfall to ENSO events and reported that during El Niño (La Niña) events, SL has less (more) rainfall in the Southwest monsoon season (June to September), while in contrast, the second inter-monsoon season (October to November) experiences more (less) rainfall than average. They also found no clear signal between ENSO events and rainfall during the first inter-monsoon season (March to May) and the Northeast monsoon season (December to February).
Other studies also noted that the ENSO impact on the rainfall variation over SL and the surrounding region could be seasonally dependent, but the arguments related to the ENSO impact on the springtime rainfall in SL are inconsistent (Vialard et al., 2011;Zubair et al., 2008). For example, Zubair et al. (2008) found that El Niño years tend to bring drier conditions in July to August and January to March periods and wetter conditions during October to December, while reporting in contrast conditions during La Niña years in SL, except for the January to March (in which showed drier conditions during La Niña events). Vialard et al. (2011) also found similar results where El Niño events induced a decrease in rainfall in the south Indian region (SL included) from January to April. However, different from Zubair et al. (2008), who suggested direr conditions in January to March periods during La Niña events, Vialard et al. (2011) noted that La Niña events contribute to the increase of January to April rainfall over SL. Therefore, it is possible that the differences between the findings of Zubair et al. (2008) and Vialard et al. (2011) are caused by the focused months are different between them. However, from these documented studies, it is unclear what is the relationship (positive or negative) between ENSO and the interannual variation of springtime (March and April) rainfall in SL. Moreover, all these documented studies focused on the daily or monthly timescale rainfall variation rather than the ENSO's impact on the diurnal rainfall activities.
This study aims to explore the possible modulations of ENSO on the SDR variability in SL and understand the underlying physical mechanisms. Exploring this issue can provide crucial information on the remote impact of ENSO on the local diurnal rainfall variation over the Indian monsoon lead region. The remaining structure of this paper is as follows: data and methods are presented in Section 2, results on rainfall analysis are in Section 3, the linkage between ENSO and diurnal rainfall is given in Section 4, and corresponding discussions are presented in Section 5. Finally, the conclusions are in Section 6.

Data and Methods
The precipitation data used in this study are from the latest version 7 of the TRMM Multi-satellite Precipitation Analysis 3B42 (henceforth TMPA.v7) (Huffman et al., 2007), and the Final run data of Integrated Multi-satellitE Retrievals for GPM version 6 (henceforth IMERG.v6) (Huffman et al., 2020). Both datasets are satellite-gauge products that have been calibrated with station rainfall data. The spatial and temporal resolution of TMPA.v7 is 0.25° × 0.25° and 3-hr, respectively. The spatial and temporal resolution of IMERG.v6 is 0.1° × 0.1° and 30 min, respectively. The capability of both TMPA.v7 and IMERG.v6 have been validated by earlier studies against station observations and shown reasonable correlations for the rainfall variations over SL (Bandara et al., 2021;Huang et al., 2023;Perera et al., 2022;Yoshimoto & Amarnath, 2017). Due to the end of TMPA.v7 in 2019 and the availability of IMERG.v6 data starting in 2001, this study focused on the period from 2001 to 2019 for all analyses.
The wind circulation and specific humidity were obtained from the fifth generation of the European Centre for Medium-range Weather Forecast reanalysis data set (ERA5) (Hersbach et al., 2020). The respective spatial and temporal resolution of ERA5 is 0.25° × 0.25° and 1-hr. Using these meteorological variables, we analyzed the vertically integrated moisture flux (MF) transportations to identify the maintenance mechanisms of rainfall variability. Following the work of Chen et al. (1988) and Huang et al. (2016), we calculated MF and the related streamfunction (Ψ MF ) and potential function ( MF ) components of MF, using Equation 1 to Equation 4.
In Equations 1-4, g, V, q, and p represent the gravity, wind, specific humidity, and pressure, respectively. A positive (negative) value of Ψ MF represents an anticyclonic (cyclonic) MF circulation. A positive (negative) value of MF represents the convergence (divergence) of MF. The sea surface temperature (SST) data were obtained from version 2 of the Optimum Interpolation SST data set of the National Oceanic and Atmospheric Administration (Reynolds et al., 2002). This data set has a 1.0° × 1.0° spatial and monthly temporal resolution. Using this data, we calculated the SST Anomaly (SSTA; climatological mean removed) indices over the NINO3.4 region (5°N-5°S, 170°W-120°W) and NINO3 region (5°N-5°S, 150°W-90°W) respectively, for two selected seasons: the winter (December to February; DJF mean) and the following spring (March to April; MA mean), for the period from 2001 to 2019. Following earlier studies (Kousky & Higgins, 2007;Wolter & Timlin, 2011), the years with winter SSTA exceeding +0.5°C and −0.5°C are defined as El Niño and La Niña events, respectively (Table 1). In addition, the subsequent springtime is used for the respective composites of ENSO events (see notes in Table 1).
Additionally, the power spectrum analysis was utilized to examine the corresponding oscillation period of selected variables on the interannual timescale (Huang et al., 2019); more details of this method can be found in DelSole and Tippett (2022). The Empirical Orthogonal Functions (EOF) analysis is applied to the sub-daily rainfall variation  to identify the dominant spatiotemporal rainfall patterns over SL and nearby regions; more details of this method can be found in Hannachi et al. (2007). Hereafter, the climatological mean denotes the average of selected variables over the springtime of 2001-2019. The diurnal anomalies of selected variables are obtained by removing the daily mean from the sub-daily variables. Hereafter, when Local Time (LT) in SL is used, it is equal to Coordinated Universal Time (UTC) +5:30 hr.

The Analysis of Rainfall Variation
Before analyzing the interannual variation, we first looked at the climatological features of the SDR over SL and nearby regions. Figure 1a shows the SDR variability over the Indian subcontinent sector of the Asian monsoon region, estimated from the TMPA.v7 during 2001-2019. It is evident that the larger variabilities are over land areas (including SL) than over oceanic regions. Several past studies have indicated that the diurnal rainfall variability is larger over land areas due to the interaction of land-sea breeze with local topography (Dai et al., 2007;Deshpande & Goswami, 2014). The details of the local topography extending over SL are given in Figure 1b, highlighting the central mountain range in the south-central part of SL. Other studies have noted that SL's central mountains profoundly influence local rainfall (Puvaneswaran & Smithson, 1991;Suppiah & Yoshino, 1984), and diurnal variability is a prominent feature in springtime over SL (Thambyahpillay, 1954). Therefore, the larger variability shown in Figure 1a demonstrates that SL is a perfect location for studying SDR variation over the Indian monsoon lead region. Figure 2a shows the temporal evolution of SDR averaged during 2001-2019 over SL. To clarify if IMERG.v6 exhibits the same evolution as seen with TMPA.v7, we used 3-hourly data from both datasets here. It is clear from Figure 2a that SDR has a distinct single peak in the late afternoon at 17:30 LT (12 UTC) in both TMPA.v7 (blue bar) and IMERG.v6 (red line). This timing of maximum rainfall at 17:30 LT (12 UTC) is consistent with the hourly IMERG.v6 data as well (shown in Figure S1 in Supporting Information S1). To further analyze the spatial consistency of this identified pattern, we constructed a corresponding spatial distribution of phase diagrams to illustrate the occurrence timing of maximum SDR in SL extracted from TMPA.v7 ( Figure 2b) and IMERG.v6 ( Figure 2c). It is evident that the afternoon maxima at 17:30 LT (12 UTC) are a common feature over most of SL in both datasets. Based on this finding, we chose only IMERG.v6 (i.e., the data set with the most recent and  highest spatiotemporal resolution) for the remainder of the analysis to reduce the repeated analysis and make this study more concise. For information on the comparison between IMERG.v6 data and rain gauge stations in illustrating the diurnal rainfall variation over SL, please refer to Huang et al. (2023).
It is important to note that Figure 2 only displays the climatological mean feature and does not provide information on the interannual variation of the diurnal rainfall cycle over SL. To elucidate the major spatiotemporal characteristics of rainfall variation at multiple timescales, previous studies have often employed EOF analysis (Hannachi et al., 2007;Huang et al., 2021;Kikuchi & Wang, 2008). Therefore, to determine whether the diurnal cycle is the dominant mode of rainfall variation over SL, we applied the EOF analysis on hourly SDR extracted from IMERG.v6 for 19 years from 2001 to 2019. Figure 3 shows the first two EOF modes, with a sample size of 456-time points (=24 hr × 19 years). The first EOF mode (EOF1) and the second EOF mode (EOF2) represent 53% and 11% of the total variability, respectively, suggesting that EOF1 is the major variation mode. The spatial distribution of the EOF1 (Figure 3a) is aligned with the spatial distribution of SDR variability revealed in Figure 1a, with a larger variability over the western than eastern SL. Additionally, the temporal variability of EOF1 ( Figure 3b) consists of a regular signal, with peaks occurring every year that can be linked to the diurnal cycle of rainfall. Different from EOF1, the EOF2 has a more prominent spatial variability over oceanic areas than on land (Figure 3c), and its temporal variability does not consist of regular peaks in every year ( Figure 3d). These EOF results imply that the diurnal variation (i.e., EOF1) is the most apparent feature seen in the springtime of 2001-2019.
Also noted from Figure 3b, the amplitude of SDR in SL is inconsistent throughout all the examined years of 2001-2019, suggesting that there is an interannual variation of SDR's amplitude in SL. To further clarify this feature, we constructed the area-averaged hourly SDR in SL for each year from 2001 to 2019 ( Figure 4a). It is evident from this year-to-year hourly rainfall variability that there is no significant variability in the phase of the hourly rainfall cycle over SL during this period. The hourly rainfall does, however, exhibit long-term interannual variability in terms of amplitude. Notably, the hourly rainfall evolution in Figure 4a can be separated into two different components: daily mean and diurnal anomalies (i.e., daily mean removed). Therefore, to better clarify the signal of diurnal variation, we also constructed a similar analysis for diurnal anomalies of rainfall area-averaged in SL ( Figure 4b). As noted from Figure 4b, all years have a period of positive anomalies of SDR, starting at around 14:30 LT (9 UTC) and ending at around 22:30 LT (17 UTC). The maximum peak of SDR appears around 17:30-18:30 LT (12-13 UTC) in all examined years, but the amplitude of SDR is larger in some years than others.
The interannual variation of the area-averaged time series of daily mean rainfall in springtime ( Figure 4c) and the amplitude of SDR (Figure 4d) over SL are further examined. For both Figures 4c and 4d, the highest value in the time series was observed in 2008, while the lowest was in 2019. Visually, there seems to be an interannual variability with roughly ∼2-4 years of oscillation. More evidence supporting this finding will be given later. In addition to the interannual variation signal, there is an overall declining signal, particularly from 2008 onwards; this feature is also seen in both the daily mean ( Figure 4c) and the SDR's amplitude ( Figure 4d). However, the time period used in this analysis (due to data availability) is too short of drawing any reliable conclusions about long-term trends in rainfall over SL. Therefore, no further detailed analysis of the long-term trends is discussed in this study. It is suggested that future studies be conducted using longer time periods of hourly rainfall data to better understand the long-term trends in rainfall over SL.
In a summary of Figures 2-4, our results indicate the following points; (a) there is strong diurnal rainfall variability over the land area over SL with a prominent afternoon rainfall peak, (b) there is no obvious change in the phase of hourly SDR cycle, but there is a notable interannual variation in the amplitude of SDR as well as the springtime daily mean rainfall. As indicated earlier, several past studies have identified the effect of ENSO events on the interannual variation of monthly rainfall in SL in different months and seasons (Chandrasekara et al., 2017;Kane, 1998;Vialard et al., 2011;Zubair & Ropelewski, 2006;Zubair et al., 2008). As the variation of the amplitude of SDR seen in Figure 4d is similar to the variation of springtime daily mean rainfall seen in Figure 4c, it is very likely that ENSO plays a role not only in regulating the seasonal mean rainfall but also the amplitude of SDR in SL. In the following section, we will examine this possibility.

The Linkage Between ENSO and Diurnal Rainfall in SL
To investigate the potential relationship between ENSO and the springtime rainfall over SL, we analyzed the spatial extent of the temporal correlation between the springtime SSTA and two selected time series: (a) the daily mean of springtime rainfall over SL (  Figure 5b that the interannual variation of SDR's amplitude over SL also negatively correlates with the springtime SSTA changes over the entire NINO3.4 region and part of the NINO3 region. However, compared to the daily rainfall correlation results in Figure 5a, the areas with significant negative correlation are expanding more westward toward the central Pacific region along the tropical region in Figure 5a than in Figure 5b. Additionally, the significant spatial correlation regions over the Indian Ocean show a notable reduction in Figure 5b compared to Figure 5a. All of these features suggest that ENSO might play a dominant role in regulating the interannual variation of SDR's amplitude over SL. Indeed, as seen in Figures 5c and 5d, both the daily mean of springtime rainfall (black line in Figure 5c) and the amplitude of SDR (black line in Figure 5d) over SL have a negative relationship with the variation of springtime SSTA over the NINO3.4 domain (red line in Figures 5c and 5d) and NINO3 domain (purple line in Figures 5c and 5d) on the interannual timescale. Through the calculation of correlation coefficients between the time series of SSTA and rainfall indices displayed in Figures 5c and 5d, the results further demonstrate that the observed negative relationship has passed the 95% significant t-test (refer to Table S1 in Supporting Information S1). This implies that when the springtime SST over the central and eastern tropical Pacific Ocean becomes cooler (warmer), the springtime daily mean rainfall and diurnal amplitude over SL increase (decrease).
Considering that the memory of SSTA might continue the changes from winter to the following springtime (Shi et al., 2022), it is possible that the SSTA in the winter can have an impact on the subsequent rainfall variation over SL as well. To clarify this assumption, we conducted an analysis similar to Figure 5 but using the previous winter SSTA to lagged-correlate with the following springtime rainfall index in SL. The result shown in Figure 6 confirms that there is also a negative relationship between winter SSTA over NINO3.4 (or NINO3) domain and the springtime rainfall variation over SL. This is true for both the interannual variation of springtime daily mean (Figures 6a and 6c) and the SDR's amplitude (Figures 6b and 6d). The main difference between Figures 5a and 6a is that the negative correlation pattern extends eastward to cover the entire NINO3 domain in Figure 6a but not in Figure 5a. Additionally, the areas showing a negative correlation pattern over the Indian Ocean in Figure 5a expand in size compared to the similar areas seen in Figure 6a. A similar difference is revealed between the comparison of Figures 6b and 5b. Despite these differences, it can be implied from Figures 6c and 6d that when an El Niño (La Niña) event occurs in winter, the following springtime rainfall over SL might be decreased (increased) and the SDR's amplitude might be suppressed (enhanced). Notably, Figures 6c and 6d also highlight that the 2008 La Niña event produced more rainfall over SL than other La Niña events. Further discussions on the underlying cause of this feature will be provided in Section 5, where we will examine the maintenance mechanism of rainfall formation over SL.
Further evidence of the close relationship between winter ENSO events and the following springtime rainfall variation over SL is presented in Figure 7, which shows the power spectrum of the indices obtained from Figures 6c and 6d. It is interesting to note that all the analyzed indices, including the springtime rainfall indices (mean rainfall and diurnal rainfall) and winter SSTA indices (NINO3.4 and NINO3), exhibit similar 2 to 4-year oscillation periods (highlighted in the yellow shaded area in Figure 7). Based on the findings of the similar oscillation periods (Figure 7) and the time-lagged relationship (Figure 6), it is suggested that the winter SSTA index over NINO3.4 and NINO3 regions can be useful for predicting the following springtime rainfall (including daily mean and diurnal amplitude) over SL.
The above suggestion can be further confirmed by the comparison between composites of springtime daily mean rainfall and the anomalies of SDR over SL for three different situations: climatological mean (Figure 8a), El Niño (Figure 8b), and La Niña (Figure 8c) years. During El Niño years, daily mean rainfall rates over SL showed less than normal, while in contrast, during La Niña years showed greater than normal rainfall (see left panel of Figure 8). Similarly, during El Niño and La Niña years, there is a decrease and increase in the SDR's amplitude, respectively, as compared to the climatological mean (see right panel of Figure 8). The percentage increase in daily rainfall and SDR amplitude during La Niña years compared to El Niño years over SL is 29.1% and 21.3%, respectively (see Table S2 in Supporting Information S1 for details on the calculation method). This further demonstrates that the daily rainfall and SDR amplitude over SL increase relatively more during La Niña years than during El Niño years. Notably, in all examined situations (Climatology, El Niño and La Niña), the formation of afternoon diurnal rainfall is regulated by the low-level circulation changes with diurnal varying land-sea breezes, which provides more convergence in the afternoon over SL than in other hours (see vectors in the right panel of Figure 8). In the following section, we will examine how large-and local-circulation patterns related to ENSO events can affect the springtime rainfall variation over SL. Figure 9a shows the climatological mean of springtime wind circulation at 850 hPa as well as the previous winter SST. Visually, SL is located over the warmer areas modulated by an east-west extension of cyclonic circulation (marked by a black dashed line in Figure 9a), which suggests the potential for active convection in this region during the springtime. The ENSO influence was analyzed by looking into the anomaly composites of El Niño and La Niña (Figures 9b and 9c calculated by removing the climatological mean). Results show that in association with an El Niño (La Niña) event, the east-west extension of cyclonic anomalies circulation over the domain of (60°E−150°E, 30°S-30°N) moves southward (northward) compared to the climatological mean feature. As a result, SL comes under the influence of the anticyclonic (cyclonic) region in springtime following a winter El Niño (La Niña) event. As the anticyclonic (cyclonic) circulation is generally unfavorable (favorable) for the development of convective activities, the suppressed (enhanced) springtime mean rainfall that we observe in El Niño (La Niña) composites in Figure 8 could be supported by this large-scale circulation influences over SL given in Figure 9.

Discussions on the Large-Scale Circulation Modulations
In addition to the circulation change, we also examined the large-scale moisture flux transport patterns related to ENSO. Following the composites method of Figure 9, we constructed composites of Ψ MF (Figures 10a-10c) and MF (Figures 10d-10f) for the climatological mean, El Niño minus the climatological mean, and La Niña minus the climatological mean, respectively. Focusing on Ψ MF , an area with negative Ψ MF (i.e., cyclonic moisture flux circulation) persists over the regions of SL and India in the springtime climatology.
Relative to the climatology, there are positive (negative) values of Ψ MF over SL and nearby regions in Figure 10b ( Figure 10c), implying that the cyclonic moisture flux circulation in the climatology is suppressed (enhanced) during El Niño (La Niña) years over SL and nearby regions. These changes in Ψ MF echo the features observed in Figure 9. Earlier, we mentioned that the anticyclonic (cyclonic) circulation patterns would hinder (facilitate) the convective development activity over SL. More direct evidence to support this discussion is revealed in Figures 10d-10f using MF . Relative to the climatology, the region over SL and nearby area is dominated by negative (positive) values of MF during El Niño (La Niña) years. This indicates that the moisture flux convergence in the region over SL becomes weaker (stronger) during El Niño (La Niña) years. As a result, local rainfall over SL decreased (increased) during El Niño (La Niña) years due to the decrease (increase) in moisture supply.
Notably, Figures 9 and 10 are constructed based on the winter NINO3.4 index. Considering that the NINO3 also shows a clear negative correlation to the SDR's amplitude (see Figure 6b), we further examined the changes in springtime Ψ MF (Figure 11b) and MF (Figure 11c) that related to the winter NINO3 index (Figure 11a), using the time-lagged correlation method. In Figure 11b, positive temporal correlation coefficient values are revealed over SL, indicating an influence from the anticyclonic (cyclonic) moisture flux circulation over SL during the El Niño (La Niña) years. In conjunction with these circulation changes, one can infer from Figure 11c that when the winter NINO3 index is positive (negative), the following springtime MF over the SL region becomes negative (positive), implying a divergence (convergence) of moisture flux supply around the SL area. Despite using different SSTA indices (NINO3.4 and NINO3), these changes in springtime Ψ MF (Figure 11b) and MF (Figure 11c) are consistent with the changes we observed in Figures 10d-10f. This supports the robustness of our findings.  Table 1. The times are given in local time (i.e., 05:30 LT = 00UTC). In the right panel, the diurnal anomalies of surface wind vectors at 10 m are added to illustrate the local surface convergence/divergence.
To further clarify how these moisture circulation changes can impact the SDR's amplitude, we constructed a similar temporal correlation analysis using the index of SDR's amplitude over SL (Figure 11d) to correlate the springtime Ψ MF (Figure 11e) and MF (Figure 11f), respectively. From Figures 11e and 11f, it can be inferred that generally, the larger (smaller) amplitudes of SDR occur when there is an intensification (suppression) of cyclonic moisture flux (i.e., Ψ MF < 0) coupled with an enhancement (suppression) of moisture flux convergence (i.e., MF > 0) over SL, resulting in an enhancement of the SDR's amplitude. Furthermore, by comparing Figures 11e and 11f and Figures 11b and 11c, we noted that the two patterns are opposites, suggesting that the moisture flux circulation patterns related to El Niño (La Niña) periods are unfavorable (favorable) for the SDR's amplitude enhancement over SL.
Further evidence to support the modulation of ENSO on the SDR's amplitude over SL is given in Figure 12, which shows the composites of selected variables during El Niño and La Niña years, respectively. As shown in Figure 12a, the SDR has a larger (smaller) amplitude in SL during La Niña (El Niño) years. To support this observed rainfall change, during different ENSO periods, we used the temporal variation of vertically integrated moisture flux convergence (denoted as −∇⋅MF) area-averaged over SL (Figure 12b), which shows an increase in the convergence of moisture flux (−∇⋅MF > 0) starting from around 11:30 LT and continuing until around 17:30 LT. The time-lagged relationship between the maximum (−∇⋅MF) and the maximum rainfall over SL that  Table 1. is shown in Figures 12a and 12b has been reported and explained in detail in Huang et al. (2023). According to Huang et al. (2023), the surface temperature in SL reaches its maximum during the daytime at around 13:30 LT. The resulting land-sea thermal contrast induces associated upward motion and moisture flux convergence, which peaks at around 15:30 LT, and then transports more moisture flux from the ocean to SL to support the maximum rainfall formation around 17:30 LT. Consistent with Huang et al. (2023), we also observed similar features in both El Niño and La Niña years for the above-mentioned time-lagged relationship between the diurnal rainfall evolution and the diurnal cycle of humidity and vertical motion (refer to supplementary Figures S2a and S2b in Supporting Information S1).
Notably, the daytime moisture flux convergence in La Niña (blue line in Figure 12b) is overall larger than that in El Niño (red line in Figure 12b) composites. Such an accumulated moisture flux convergence can cause a delay in the peak of SDR amplitudes, which occurs at 17:30 LT and is stronger during La Niña years than El Niño years (as shown in Figure 12a). To further support this idea, we compared the El Niño ( Figure 12c) and La Niña (Figure 12d) composites by analyzing the accumulated moisture flux and related SDR during the time period from 14:30 to 17:30 LT, after removing the climatological mean values. Results shown in Figures 12c and 12d confirm that the decrease (increase) in daytime moisture supply during El Niño (La Niña) years supports a decrease (increase) in SDR's amplitude over SL. Moreover, the examinations on the vertical motion and humidity provided in supplementary Figure S2c ( Figure S2d in Supporting Information S1) also revealed suppressed (enhanced) upward motion and humidity during El Niño (La Niña) years, which further confirms the ENSO modulation of SDR in SL.
Finally, based on the maintenance mechanisms of rainfall formation suggested in the analysis of Figures 10-12, we explained why the 2008 La Niña event exhibited a more significant increase in springtime rainfall and SDR over SL than other La Niña years. As shown in supplementary Figure S3 in Supporting Information S1, the 2008 La Niña event had colder SSTA distribution over the eastern tropical Pacific Ocean compared to the average of all  Table 1.
La Niña events. The colder SST expanded over the region between 150°E-120°W, leading to a divergence center over 170°W-140°W in the 2008 La Niña event. This divergence center was west of the average divergence center of all La Niña events, over 150°W-120°W. The westward shifts of the divergence center in the eastern tropical Pacific Ocean imply a westward shift of the tropical east-west Walker circulation, which can lead to a westward shift of the convergence center in the western tropical Pacific Ocean. As demonstrated in Figure S3a in Supporting Information S1, the average convergence center (denoted by the yellow star in Figure S3 in Supporting Information S1) of all La Niña events was located over the Maritime continent and shifted westward to western Indonesia in the 2008 La Niña event. As a result, the convergence over SL was stronger in the 2008 La Niña event than in other La Niña events. Consequently, a similar westward shift of the moisture flux convergence center appeared in the 2008 La Niña event relative to other La Niña events (see Figure S4 in Supporting Information S1). As a result of the enhanced moisture supply over SL, the increase in rainfall over SL was more significant in the 2008 La Niña event than in the average of all La Niña events. This further supports the findings revealed in Figures 10-12 regarding the role of convergence of moisture flux changes in maintaining the rainfall changes over SL.

Conclusions
In this study, we analyzed the interannual variability of SDR and explored the potential role of ENSO events in regulating this variability. By analyzing high-spatiotemporal resolution satellite precipitation data from 2001 to 2019, we found that the SDR in SL featured a clear single peak in the late afternoon at 17:30 LT (12 UTC) and has relatively consistent temporal phase evolution across the Island with minimal spatial variability (see Figures 1  and 2). By applying EOF analysis to springtime hourly rainfall data from 2001 to 2019, we found that most of the observed spatiotemporal variability in springtime rainfall is due to diurnal variation and that the SDR's amplitude has a notable interannual variability (see Figure 3); this interannual variation is further demonstrated in Figure 4. Further analysis revealed that the interannual variation of the SDR's amplitude and springtime daily mean rainfall in SL are both negatively correlated with the springtime SSTA over the central and eastern tropical Pacific Oceanic region, indicating a potential influence of ENSO on springtime rainfall variability in SL (see Figure 5). Moreover, we found a potential for using previous winter SSTA indices (for both NINO3 and NINO3.4 regions) to predict the following SDR activity in SL (see Figure 6).
It is noted that when a La Niña (El Niño) event occurs during the winter, the following SDR activity over SL increases (decreased) (see Figure 7). Further power spectrum analysis confirms the relationship between SDR in SL and ENSO events, as both exhibit common oscillation periods of 2 to 4-year (see Figure 8). In examining the associated circulation changes (see Figures 9-11), we found that in La Niña (El Niño) years, there is an enhanced (suppressed) cyclonic circulation and enhanced (suppressed) moisture flux convergence over SL and the surrounding region. These large-scale wind and moisture circulation changes lead to an increase (decrease) in springtime local convective rainfall potential in SL by modulating moisture convergence (divergence) over SL. Moreover, we found that the local-scale daytime moisture flux convergence in SL is generally larger in La Niña years compared to El Niño years; this provides a larger moisture supply for the enhancement of SDR's amplitude in SL in La Niña than in El Niño years (see Figure 12).  Table 1.
To the best of our knowledge, this is the first study to examine the interannual variability of SDR characteristics and its potential causes in SL. In contrast to previous studies that focused mostly on the interannual variation of seasonal mean rainfall, we have moved a step forward to demonstrate the influence of ENSO on the SDR in SL and demonstrated how the related atmospheric circulation changes could affect the SDR changes. These findings are important to understand how the ENSO can impact the local rainfall features over the Indian summer monsoon lead region. Finally, we would also like to highlight the importance of studying climate trends of the SDR over this region (e.g., features seen in Figures 4c and 4d) when adequate lengths of satellite precipitation estimate data become available in the future.